In a typical microcircuit there are seven layers of patterning. The size of a microcircuit is determined by the minimum possible line width and line spacings which can be obtained. These in turn are, in part determined by the accuracy with which one pattern can be placed on top of another. The smaller the error in overlay placement the greater the packing density which can be achieved. Typically, machines now being made for projection lithography have specifications of the order of 1/2 micron. This error is made up of two factors. One is an alignment error and the other is a distortion error. Alignment error is determined by the ability of the operator or automatic alignment system of the machine to position one pattern level over the another. Distortion is the error inherent in the imaging system of the machine. It is desired to be able to measure both types of error and to maintain accuracy within the prescribed limit with respect to both.
It is necessary that overlay accuracy measurements be made during production and initial calibration of machines and also periodically in service to assure that accuracy is being maintained. One way that this has been done in the past was to print vernier patterns on a substrate and optically to read these patterns. For example, a pattern consisting of a set of lines with a pitch of 10 microns would be placed next to another pattern with a pitch of 9.75 microns. If overlay were perfect, the two sets of lines would line up at their center and, going out from the center, each line of the second pattern would be progressively closer to the center than the corresponding line of the first pattern. If overlay were in error by 0.25 .mu.m, then the two sets of lines would line up at a position displaced by one line from the pattern center.
Furthermore, in carrying out the measurements of this nature it should be noted that measurements are taken at a plurality of sites on a wafer and that checks are made in both the X and Y direction.
The main problem with optical measurements, other than the fact that they depend on human operators, is that they are limited in accuracy. In the above example, it is only possible to read to a quarter of a micron and to interpolate to an eighth of a micron. Thus, there have been proposals to utilize electrical techniques. The advantage of electrical techniques is that they permit automated and more accurate measurements utilizing probes at appropriate stations. Examples of literature describing prior art electrical measurement techniques are the following:
Solid State Technology, "Microelectronic Test Structures for Characterizing Fine-Line Lithography", D. S. Perloff, T. F. Hasan, D. H. Hwang and J. Frey, May, 1981, pp. 126-129 and 140. PA1 Solid State Technology, "Use of Microelectronic Test Structures to Characterize IC Materials, Processes, and Processing Equipment,", G. P. Carver, L. W. Linholm and T. J. Russell, September, 1980, pp. 85-92. PA1 Solid-State Technology, "Real-Time Monitoring of Semiconductor Process Uniformity," D. S. Perloff, T. F. Hasan and E. R. Blome, February, 1980, pp. 81-86. PA1 Solid State Science and Technology, "Alignment, and Mask Errors in IC Processing," K. H. Nicholas, I. J. Stemp and H. E. Brockman, March, 1981, pp. 609-614. PA1 Fifteenth Symposium on Electron, Ion and Phonton Beam Technology, "Performance Limits in 1:1 UV Projection Lithography," J. H. Bruning, May, 1979, pp. 1-8.
However, in the resistance measurements which are known in the prior art, first and second levels of a conductor are constructed such as to obtain a pair of resistors which, if the overlay is perfect will have equal values. The difference in resistance value is a measure of the offset error. However, using these prior art techniques, the first or base level was destroyed, by the overlay of the second layer, each time a test was carried out. This required having a large plurality of base patterns constructed for testing. However, it is difficult reproducibly and accurately to make base patterns. The base pattern must be as distortion-free as possible. This is normally accomplished by contact printing. However, there is a limit to the number of circuits which can be contact printed with a master mask. Thus, the ideal would be a base pattern which could be used over and over again for calibration, maintenance and so forth. With such a base pattern, since the same base pattern would be used from time to time, any changes would be caused by changes in the apparatus being tested and could not be attributed to a new base pattern. Over a period of time, in addition to saving time and money, the availability of a reusable base pattern would insure better accuracy and repeatability.